Face Tells Detailed Expression

Overview

This repository provides an implementation of paper, Face Tells Detailed Expression: Generating Comprehensive Facial Expression Sentence through Facial Action Units.

Prerequisites

• Python 2.7 (code has been tested with this version)
• Download vgg_face_weights and put it in the project.
• Download pycocoevalcap and put it after unzipping to here.

Datasets

• Text-based dataset with comprehensive facial expression sentence is available here.
• We provide the code that is executable in CK+ dataset.
• We have cooked the dataset based on the format below. We provide several examples in here.

CK+ dataset

data_root 
├── Dataset/CK+ (or any other speaker-specific folder)
|	├── cohn-kanade-images/ (will contain the aligned video image frames)
|	├── FACS/		(will contain the facial action units info.)
|	├── Emotion/	(will contain emotion info.) 

Citation

Please cite the following paper if you have use this code:

@inproceedings{hong2020face,
  title={Face Tells Detailed Expression: Generating Comprehensive Facial Expression Sentence Through Facial Action Units},
  author={Hong, Joanna and Lee, Hong Joo and Kim, Yelin and Ro, Yong Man},
  booktitle={International Conference on Multimedia Modeling},
  pages={100--111},
  year={2020},
  organization={Springer}
}